Objectives: To identify the etiologies and risk factors associated with fever of nosocomial origin and compare the outcomes of patients with and without nosocomial fever.
Design: A prospective study with a case-control component.
Setting: The three medical services of a university hospital.
Patients and methods: We evaluated 100 patients with fever of nosocomial origin. Etiologies were assigned by the consensus of three independent reviewers using Centers for Disease Control guidelines. Predictors of bacterial etiology were identified using logistic regression methods. Controls matched for age (+/- 5 years), sex, and comorbidity were identified for 65 of the nosocomial fever cases. Cases and controls were compared with conditional logistic regression to identify risk factors for fever of nosocomial origin. Cases and controls were followed at 1 month to compare length of stay, mortality, and hospital readmission rates.
Measurements and main results: An etiology was assigned in 81 cases: 51 had bacterial infection, 5 had nonbacterial infection, and 25 had a noninfectious etiology. Multivariate predictors of bacterial infection were the presence of diabetes mellitus (odds ratio [OR] = 5.74; 95% confidence interval of the odds ratio [CI] = 1.61 to 20.40), length of stay to fever onset after admission greater than 10 days (OR = 3.98, CI = 1.04 to 15.22), maximum temperature greater than 38.7 degrees C (OR = 3.37, CI = 1.28 to 8.88), and white blood cell count greater than 10 x 10(9)/L (OR = 4.64, CI = 1.68 to 12.77). Comparison of cases and controls indicated that patients with nosocomial fever had more invasive procedures in the 72 hours preceding enrollment in the study than controls (OR = 1.46, CI = 1.03 to 2.08). Although cases were hospitalized almost twice as long as controls (21.5 versus 12.5 days; p < 0.0001), neither 30-day mortality nor hospital readmission rates differed between cases and controls.
Conclusion: Nosocomial fever often does not represent nosocomial infection and may constitute a wide variety of other conditions. Since bacterial etiology of fever is a major concern, clinical and laboratory findings can be used to predict patients at high risk for an infectious process. A reduction in the number of invasive procedures may decrease the incidence of nosocomial fever. New-onset nosocomial fever leads to an increase in length of hospital stay but is not a significant predictor of mortality.